Research on the Application of E-Mail Classification Based on Support Vector Machine

Abstract

The volume of junk E-mail on the Internet has grown tremendously in the past few years, and this problem has attractes many researchers attention. Due to the diversity of E-mails and their growing dimensionalists traditional methods of classification are slow speed and inaccurate. In order to improve the accuracy of classification, an E-mail classification method is proposed based on support vector machine. E-mail classification tasks consist of feature extraction and classification. Mutual information method is used to extract key features of E-mail while support vector machine is designed to classify. Simulation experiments on nine classes E-mails, support vector machine’s average accurates is 89.9%. In comparison with BPNN method, the classification performances are improved by 4%. Experimental results indicate that support vector machine is useful method for E-mail classification.